1 / 39

Bandwidth Aware Peer-to-Peer 3D Streaming NetGames 2009

Bandwidth Aware Peer-to-Peer 3D Streaming NetGames 2009. Chien-Hao Chien , Shun- Yun Hu , Jehn-Ruey Jiang Adaptive Computing and Networking (ACN) Laboratory Department of Computer Science and Information Engineering National Central University, Taiwan. In a Nutshell.

kalei
Télécharger la présentation

Bandwidth Aware Peer-to-Peer 3D Streaming NetGames 2009

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Bandwidth Aware Peer-to-Peer 3D StreamingNetGames 2009 Chien-HaoChien, Shun-YunHu, Jehn-Ruey Jiang Adaptive Computing and Networking (ACN) Laboratory Department of Computer Science and Information Engineering National Central University, Taiwan

  2. In a Nutshell • We have proposed BASP, a Bandwidth Aware Peer Selection scheme that improves peer-to-peer (P2P) 3D streaming in networked virtual environments (NVEs) with the help of • broadened data sources • bandwidth reservation • tit-for-tat Adaptive Computing and Networking Laboratory Lab

  3. Outline Introduction Goals Proposed Scheme Evaluation Conclusion National Central University, Taiwan

  4. Networked Virtual Environments (NVEs) NVEs are computer-generated, synthetic virtual worlds with 3D content. Users may interact with each other in NVE via network connections. National Central University, Taiwan

  5. Two Trends in Virtual Environments (VEs) Adaptive Computing and Networking Laboratory Lab • Larger and more dynamic content • More worlds • For example, in Second Life • there are 37TB 3D content data • there are 14,150 regions in April, 2008. • The 3D streaming technique arises due to this trend.

  6. Types of NVE Content Distribution • Complete Installation • Users acquire and install all content before rendering • World of Warcraft (WoW): 8 GB • 3D Streaming • Users progressively download 3D content of objects within an area of interest (AOI) when rendering • Second Life: First Installation 22MB National Central University, Taiwan

  7. Progressively Downloading Refinements Base 1 2 3 (Hoppe 96) User • Model meshes are fragmented into base & refinements • Rendering can start without a full download of an object’s data • The more are the data, the finer is the rendering National Central University, Taiwan

  8. Area of Interest (AOI) Adaptive Computing and Networking Laboratory Lab

  9. NVE Content Requesting Models National Central University, Taiwan • Client/Server • All requests are sent to the server or server cluster • Peer-to-Peer (P2P) • Requests can be sent to peers and the server

  10. C/S vs. P2P New object notification Request 3D content from other peers Request 3D content from the server New object notification Request 3D content from the server 1 2 3 2 Server User National Central University, Taiwan

  11. P2P 3D Streaming : Flowing Level of Detail (FLoD) [INFOCOM 2008][IEEE IC] triangles: neighbors rectangles: objects • VE is partitioned into cells with scene descriptions • AOI neighbor lists are provided by a P2P VON overlay • Users perform the following actions • Source Discovery • State Exchange • Source Selection • Content Exchange National Central University, Taiwan

  12. AOI Neighbor Management via VON P2P Overlay Voronoi diagrams identify boundary neighbors for neighbor discovery Non-overlapped neighbors Boundary neighbors New neighbors [Hu et al. 06] National Central University, Taiwan

  13. Actions in FLoD • Source Discovery • Users send queries to AOI neighbors for discovering necessary data • State Exchange • The list of available data is exchanged passively • Source Selection • Users randomly select available data • Content Exchange • First come first serve National Central University, Taiwan

  14. Problems of FLoD • Since source discovery is confined to AOI neighbors, other potential peers with necessary data may be ignored. • Since the state of available data is exchanged passively, it is not efficient. (One of our early papers has proposed exchanging the state proactively.) • Since source selectionis random and content exchange is FCFS, bandwidth utilization may be low and latency may be long. National Central University, Taiwan

  15. Outline Introduction Goals Proposed Scheme Evaluation Conclusion National Central University, Taiwan

  16. Goals Exploiting all possible content resources Increasing bandwidth utilization Reducing latency National Central University, Taiwan

  17. Outline Introduction Goals Proposed Scheme Evaluation Conclusion National Central University, Taiwan

  18. Bandwidth Aware P2P 3D Streaming • Broadened Source Discovery • A user discovers available data sources from AOI neighbors and peers in the peer list (provided by the server) • Bandwidth Reservation • Bandwidth is allocated to “good” peers • Dual-Order Content Exchange • Two order for content exchange National Central University, Taiwan

  19. Broadened Source Discovery Scene description request Description and peer list • AOI neighbors • Provided by P2P Overlay • Peer list peers • Provided by the server when a user requests a new scene description or when it explicitly requests them due to the lack of sources National Central University, Taiwan

  20. Proactive State Exchange and Bandwidth Revervation Allocated bandwidth of connection channels for “good” peers Bandwidth reserved for AOI neighbors for exchanging states and for downloading Object lists are exchanged proactively and incrementally Connection channels of fixed bandwidth are reserved for “good” peers National Central University, Taiwan

  21. What are good peers? Tit-for-Tat Strategy:Those providing more data are good peers Good peers are chosen from AOI neighbors and from peers in the peer list A peer constructs connection for good peers, called connection neighbors, and reserves a fixed-bandwidth channel to each of them. National Central University, Taiwan

  22. Dual-Order Content Exchange • First come first serve (FCFS) • For normal AOI neighbors • Early request first (with best effort guarantee) • Tit-for-tat (TFT) • From connection neighbors (peers) • High contribution first (with QoS guarantee) Adaptive Computing and Networking Laboratory Lab

  23. Outline Introduction Goals Proposed Scheme Evaluation Conclusion National Central University, Taiwan

  24. Simulation Environment National Central University, Taiwan

  25. System Performance Metrics • Server Request Ratio (SRR) • Ratio of data downloaded from the server • Fill ratio • Ratio of total data downloaded to the data required for a complete scene in AOI • Base Latency • Duration between requesting and obtaining the base piece National Central University, Taiwan

  26. Simulation Scenario (1) • To increase the number of objects • for evaluating bandwidth utilization • with 100 to 500 objects • and 100 peers National Central University, Taiwan

  27. Server Request Ratioand Average Fill Ratio Bandwidth Utilization National Central University, Taiwan

  28. Average Base Latency National Central University, Taiwan

  29. Simulation Scenario (2) • To increase the number of peers • for evaluating system scalability • with 50 to 450 peers • and 100 objects National Central University, Taiwan

  30. Server Request Ratio and Fill Ratio National Central University, Taiwan

  31. Average Base Latency National Central University, Taiwan

  32. Conclusion • Broadened Source Discovery • Peer list increases potential sources • Bandwidth Reservation • Channel allocation guarantees QoS • Dual-Order Content Exchange • Tit-for-Tac improves bandwidth utilization • Simulation results justify our claims National Central University, Taiwan

  33. Thank you for listening! National Central University, Taiwan

  34. Q&A National Central University, Taiwan

  35. 3D Streaming vs. Media Streaming • Video media streaming is very matured • User access patterns are different • Highly interactive Latency-sensitive • Behaviour-dependent Non-sequential National Central University, Taiwan

  36. Preparation incremental availability information Type Obj_ID Piece_ID Obj_ID Piece_ID ‧‧‧‧ • State Exchange • Peers periodically exchange incremental content availability information with AOI and connection neighbors . National Central University, Taiwan

  37. ‧ ‧ ‧ ‧ ‧ ‧ ‧ ‧ ‧ ‧ ‧ ‧ ‧ LODDT Object Tree Node Aura National Central University, Taiwan

  38. ‧ ‧ ‧ ‧ LODDT Object Tree Node Aura U National Central University, Taiwan

  39. LODDT (cont.) Candidates Requests • Discovery • Estimation • Selection • Every peer samples the time-to-serve (TTS) of its neighbors • Requestors organize their data requests so as to obtain tree nodes in the right order • Drawback: incorrect estimation, congestion National Central University, Taiwan

More Related